Dryrun pictures

Weaknesses of the model

1. No seasonality
2. Homogeneous fleet
3. One quota market for Sablefish
4. Alternative fisheries not simulated

What’s the point?

• We have a model and we have some data
• We want:
  • Calibrate
  • Select
  • Validate
  • Analyse

Calibrate

• Only a few free parameters
  • Catchability per species
  • Average hold size
  • Behavioural Parameters
• We want to change them such that model is as close as possible to real data

Calibration procedure

1. Force agents to go where logbook says they went
2. Fit catchabilities and hold size so that simulated data looks like real data
3. Now fix catchabilities and vary behavioural parameters
4. Pick behavioural parameters that minimize distance to logbook data
5. Check fully calibrated model against aggregate data again

Step 2

Step 5 - heatmapper

Select

• We have many options for heuristics
  • Bandit agents
  • Imitative agents
  • Heatmap agents
• Which heuristic is best?
• Rank them by quality of fit back to aggregate data

Rank

## # A tibble: 9 × 2
##         name     error
##        <chr>     <dbl>
## 1  annealing 167.82941
## 2     bandit  17.47207
## 3    clamped  17.03515
## 4        eei  16.98512
## 5       eei2  16.94044
## 6 intercepts  11.54279
## 7     kernel  13.79445
## 8    perfect  18.59034
## 9     random  32.21017

Validate

• Look at other indicators that were not calibrated against
• Do they look okay?

Quota Price - Sablefish

Profits

Active fishers

Analyse

Active Fishers

Profits